Work
Projects from my work at Lark Health, Dave & Buster’s, and American Diamond Logistics. The case studies describe what I built and how the teams used it.
Incomplete tracking and conflicting attribution rules left channel performance unreliable as direct mail faced discontinuation.
Findings reversed the decision to discontinue direct mail, a channel with $550K in annual spend. Unknown attribution fell from about 42% to 8% on a trailing 365-day view. Recovered attribution is a measurement correction, not causal lift.
Built and shipped one-year enrollment forecasts for Carelon, BCBSAL, and MMO using audience size, marketing sends, enrollment history, and seasonality.
Used to plan send volumes and direct-mail budgets, with email costs considered. Monthly enrollment error, checked separately for each partner, has usually been within 10% in 2026 to date, as of September.
Leadership needed a Q4 media budget recommendation within three hours and an explanation for a 42% year-over-year decline in web traffic.
Recommended $6.9M in spend against an outside proposal of about $11M. The analysis shaped CEO and CMO planning. Web analysis traced a major part of the traffic decline to weaker CTR and message resonance.
Established the pricing analytics department and built a HubSpot workflow for sales teams to request freight quotes.
Pricing response time improved about 43% and win rate about 22%. Selected and trained an internal analyst, and taught logistics coordinators to use the workflow.
Lifecycle activity and downstream program milestones were not connected in production measurement, including for Lark’s Medicare DPP launch.
Built production Snowflake and dbt models connecting Braze email, SMS, and push activity to downstream behavior and program milestones, and defined Marketing’s measurement path from paid acquisition through enrollment, activation, engagement, and billable milestones.
I’m studying airport disruption forecasts alongside marketing incrementality and budget allocation. These projects show the work as it develops.
When does weather disrupt a U.S. airport, and how early can we know? An interactive airport risk map with delay probabilities, chronological backtests, and visible uncertainty.
100 major U.S. airports assessed, with airport-specific models, transparent coverage gaps, and winter and spring backtests at 24, 12, and 6 hours.
Research into measuring marketing’s incremental impact and using those estimates to allocate budgets.
Planned; implementation has not started.